A couple of recent commits (IMPALA-4299 and IMPALA-4259) changed the
behaviour of these scripts slightly. Before ./buildall.sh indiscriminately
clobbered and regenerated test cluster config files, so running something
like the below would regenerate the test cluster configs, then restart the
./buildall.sh -notests -noclean && ./testdata/bin/run-all.sh
Those commits changed the behaviour so that the test cluster configs are
only regenerated if you ask buildall.sh to perform an action that requires
the test cluster. If you want to achieve the same effect as above, you
should instead run:
./buildall.sh -notests -noclean -start_minicluster
If you use the old worklow you may see run-all.sh fail after a clean.
This behaviour was changed for a number of reasons:
- So Impala can be built without the CDH test cluster dependencies.
- To speed up incremental builds run via buildall.sh
- So that buildall.sh doesn't have undocumented side effects
- So that buildall.sh doesn't put the test cluster in an undefined state
(it could create new configs but leave the cluster running with the old
Please let me know if you have any questions and sorry for any disruption.